Stock Price Forecast

Using Technical Analysis and Machine Learning

Part 1. Technical Analysis

This section provides a comprehensive technical analysis of stock prices using multiple indicators. For each stock, I calculate key metrics:

1. Trend Indicators:

• MA50/MA200: 50-day and 200-day Simple Moving Averages compare current price to medium/long-term trends (Price > MA = bullish +1, Price < MA = bearish -1)

• EMA20: 20-day Exponential Moving Average gives more weight to recent prices for short-term trend direction

2. Momentum Indicators:

• RSI (14-day): Measures overbought (>70 = -1) vs. oversold (<30 = +1) conditions, with 30-70 being neutral

• MACD: Signal line crossover indicator (MACD > 0 = +1 bullish momentum, MACD < 0 = -1 bearish momentum)

3. Volatility & Volume Indicators:

• Bollinger Bands: Price above upper band = overbought (-1), below lower band = oversold (+1), within bands = neutral (0)

• MFI (Money Flow Index): Volume-weighted RSI (>80 = -1 overbought, <20 = +1 oversold, 20-80 = neutral)

Each indicator is scored (+1 for bullish, -1 for bearish, 0 for neutral), with a final aggregate signal determining the overall market outlook (Positive, Negative, or Neutral).

# Load necessary libraries
library(tidyverse)
library(dplyr)
library(tidyquant)
library(TTR)
library(xgboost)
library(plotly)
library(gt)
library(gtExtras)
library(readr)

# 1. Load Data
# Assuming the same directory structure
df <- read_csv("data/cleaned/stock_prices.csv")
df$date <- as.Date(df$date)
df <- df %>% arrange(ticker, date)

# Define function to calculate indicators
calculate_indicators <- function(data) {
  # Create the HLC matrix required by TTR functions
  hlc <- data %>% select(high, low, close)
  
  data %>%
    mutate(
      SMA_50  = SMA(close, n = 50),
      SMA_200 = SMA(close, n = 200),
      EMA_20  = EMA(close, n = 20),
      RSI_14  = RSI(close, n = 14),
      # FIX: Pass HLC as one argument and volume as the second
      MFI_14  = MFI(hlc, volume = data$volume, n = 14)
    ) %>%
    # MACD returns multiple columns
    bind_cols(as_tibble(MACD(data$close, 12, 26, 9, maType="EMA"))) %>%
    # Bollinger Bands
    bind_cols(as_tibble(BBands(data$close, n = 20, sd = 2)))
}

# Apply calculations to all tickers
df_ta <- df %>%
  group_by(ticker) %>%
  filter(n() > 200) %>%
  group_modify(~ calculate_indicators(.x)) %>%
  ungroup()

# Scoring Logic
df_ta <- df_ta %>%
  mutate(
    MA50_Score  = case_when(close > SMA_50 ~ 1, close < SMA_50 ~ -1, TRUE ~ 0),
    MA200_Score = case_when(close > SMA_200 ~ 1, close < SMA_200 ~ -1, TRUE ~ 0),
    EMA_Score   = case_when(close > EMA_20 ~ 1, close < EMA_20 ~ -1, TRUE ~ 0),
    MACD_Score  = case_when(macd > signal ~ 1, macd < signal ~ -1, TRUE ~ 0),
    RSI_Score   = case_when(RSI_14 < 30 ~ 1, RSI_14 > 70 ~ -1, TRUE ~ 0),
    BB_Score    = case_when(close < dn ~ 1, close > up ~ -1, TRUE ~ 0),
    MFI_Score   = case_when(MFI_14 < 20 ~ 1, MFI_14 > 80 ~ -1, TRUE ~ 0)
  ) %>%
  mutate(
    Total_Score = MA50_Score + MA200_Score + EMA_Score + MACD_Score + RSI_Score + BB_Score + MFI_Score,
    Signal = case_when(Total_Score > 0 ~ "Positive", Total_Score < 0 ~ "Negative", TRUE ~ "Neutral")
  )

# Generate Summary Table
summary_df <- df_ta %>%
  group_by(ticker) %>%
  filter(date == max(date)) %>%
  select(ticker, close, volume, SMA_50, SMA_200, EMA_20, macd, RSI_14, up, dn, MFI_14, 
         contains("Score"), Signal)

# Display styled table using gt
summary_df %>%
  gt() %>%
  tab_header(title = "Technical Analysis Summary - Latest Signals") %>%
  fmt_number(columns = c(close, SMA_50, SMA_200, EMA_20, up, dn), decimals = 2) %>%
  fmt_number(columns = volume, suffixing = TRUE) %>%
  gt_color_rows(Total_Score, palette = c("red", "yellow", "green")) %>%
  tab_style(
    style = cell_fill(color = "#d4edda"),
    locations = cells_body(columns = Signal, rows = Signal == "Positive")
  ) %>%
  tab_style(
    style = cell_fill(color = "#f8d7da"),
    locations = cells_body(columns = Signal, rows = Signal == "Negative")
  )

Technical Analysis Summary - Latest Signals
close	volume	SMA_50	SMA_200	EMA_20	macd	RSI_14	up	dn	MFI_14	MA50_Score	MA200_Score	EMA_Score	MACD_Score	RSI_Score	BB_Score	MFI_Score	Total_Score	Signal
AAPL
248.68	20.37M	269.78	234.74	259.27	-2.36946399	23.54549	279.58	243.59	25.31303	-1	1	-1	-1	1	0	0	-1	Negative
ADBE
302.73	1.71M	331.50	357.65	318.17	-3.79778065	34.05754	369.17	281.19	32.94964	-1	-1	-1	-1	0	0	0	-4	Negative
AMZN
239.68	17.27M	232.37	220.30	236.04	0.59766662	55.47537	248.51	225.03	61.55384	1	1	1	-1	0	0	0	2	Positive
BAC
51.58	14.33M	53.90	48.30	53.79	-1.23120016	34.56166	58.15	51.02	45.30107	-1	1	-1	-1	0	0	0	-2	Negative
DIS
111.56	4.05M	109.76	110.60	112.44	0.39737151	48.28802	115.90	110.75	59.20544	1	1	-1	-1	0	0	0	0	Neutral
GOOGL
328.39	12.24M	312.72	229.67	323.91	1.80127254	58.32054	339.59	306.99	58.85994	1	1	1	-1	0	0	0	2	Positive
HD
383.07	996.84K	356.24	370.44	368.76	2.17222348	65.62475	395.07	332.06	75.39303	1	1	1	1	0	0	0	4	Positive
JNJ
218.29	2.19M	206.78	176.21	212.80	1.79453697	67.65978	222.64	199.76	60.93761	1	1	1	1	0	0	0	4	Positive
JPM
297.76	4.51M	312.95	288.98	313.03	-1.20877563	34.77893	340.64	296.11	45.10625	-1	1	-1	-1	0	0	0	-2	Negative
KO
72.26	5.75M	70.56	69.11	70.70	0.70710718	62.89778	72.89	67.56	62.38649	1	1	1	1	0	0	0	4	Positive
MA
525.59	1.74M	553.92	561.10	550.38	-1.56460317	30.93978	598.33	519.88	35.03904	-1	-1	-1	-1	0	0	0	-4	Negative
META
661.78	11.57M	639.49	676.81	640.07	-0.82320725	57.21962	683.17	605.76	56.27555	1	-1	1	1	0	0	0	2	Positive
MSFT
469.30	19.93M	480.94	482.82	468.51	-1.67824555	48.27677	498.43	445.76	42.65325	-1	-1	1	-1	0	0	0	-2	Negative
NFLX
86.00	34.57M	97.70	113.01	89.75	-3.64960133	31.64944	96.02	84.00	20.03264	-1	-1	-1	-1	0	0	0	-4	Negative
NVDA
187.94	80.23M	183.83	165.69	184.93	0.08988118	54.29832	191.68	180.92	46.66409	1	1	1	-1	0	0	0	2	Positive
PG
150.58	6.52M	144.97	153.20	145.17	0.73180086	64.85683	150.08	138.06	66.20983	1	-1	1	1	0	-1	0	1	Positive
PYPL
56.91	4.05M	60.13	67.19	57.83	-2.06688642	42.09407	60.67	55.21	35.16651	-1	-1	-1	1	0	0	0	-2	Negative
TSLA
448.26	34.98M	442.38	372.75	444.51	-0.77553181	51.28174	475.57	416.33	55.61414	1	1	1	-1	0	0	0	2	Positive
UNH
355.18	3.11M	331.31	332.55	340.37	1.29486565	63.18486	355.43	323.15	58.84521	1	1	1	1	0	0	0	4	Positive
V
327.00	2.25M	338.03	344.07	336.65	-1.43186372	35.02471	367.72	316.84	40.38805	-1	-1	-1	-1	0	0	0	-4	Negative
WMT
118.40	8.19M	112.18	102.04	116.44	1.71530489	60.19195	122.17	108.92	27.87472	1	1	1	1	0	0	0	4	Positive

Part 2. Machine Learning

1. Overview

Building upon the technical analysis in Part 1, this section utilizes Machine Learning (XGBoost) to predict stock prices. Unlike traditional indicators that give simple Buy/Sell signals, the ML model analyzes the complex relationships between historical patterns (RSI, MACD, Moving Averages) to forecast the exact closing price of the next trading day.

2. Methodology

• Algorithm: We use XGBRegressor (Extreme Gradient Boosting), a robust algorithm highly effective for structured time-series data.

• Feature Engineering: The model inputs include Open, High, Low, Volume, and all technical indicators calculated in Part 1 (SMA, EMA, Bollinger Bands, etc.).

• Training & Validation: To prevent “data leakage” (looking into the future), the data is split chronologically:

  • Training Set (First 80%): Used to teach the model historical patterns.

  • Test Set (Last 20%): Used to evaluate how well the model predicts unseen data.

3. Interactive Analysis Dashboard

The visualization below provides a comprehensive view of the model’s performance. Use the dropdown menu to select a specific ticker:

• Top Chart (Actual vs. Predicted): Compares the real market price (Blue line) against the model’s prediction (Orange dotted line). The closer the lines, the better the model accuracy.

• Bottom Chart (Feature Importance): Ranks which technical indicators were most influential in determining the price. For example, if RSI has a high bar, the model relies heavily on momentum to make predictions for that specific stock.

Note: The table below summarizes the prediction for the next upcoming trading day, including the predicted percentage change.

# Prepare features
feature_cols <- c("open", "high", "low", "close", "volume", 
                  "SMA_50", "SMA_200", "EMA_20", "RSI_14", 
                  "macd", "signal", "up", "mavg", "dn", "pctB", "MFI_14")

ml_results <- list()
ticker_preds <- list()

for (t in unique(df_ta$ticker)) {
  ticker_data <- df_ta %>% filter(ticker == t) %>% drop_na(any_of(feature_cols))
  
  # Create Target (Shifted Close)
  ticker_data$target <- lead(ticker_data$close)
  
  # Split data
  train_df <- ticker_data %>% drop_na(target)
  split_idx <- floor(0.8 * nrow(train_df))
  
  train_set <- train_df[1:split_idx, ]
  test_set  <- train_df[(split_idx + 1):nrow(train_df), ]
  
  # XGBoost Matrices
  dtrain <- xgb.DMatrix(data = as.matrix(train_set[, feature_cols]), label = train_set$target)
  dtest  <- xgb.DMatrix(data = as.matrix(test_set[, feature_cols]), label = test_set$target)
  
  # Train Model
  model <- xgboost(data = dtrain, nrounds = 100, objective = "reg:squarederror", 
                   eta = 0.05, max_depth = 5, verbose = 0)
  
  # Predict on test set
  preds <- predict(model, dtest)
  rmse_val <- sqrt(mean((test_set$target - preds)^2))
  
  # Predict Next Day
  latest_data <- tail(ticker_data, 1)
  next_pred <- predict(model, as.matrix(latest_data[, feature_cols]))
  
  ml_results[[t]] <- tibble(
    Ticker = t,
    Current_Price = latest_data$close,
    Predicted_Price = next_pred,
    Change_Pct = ((next_pred - latest_data$close) / latest_data$close) * 100,
    RMSE = rmse_val
  )
}

final_ml_results <- bind_rows(ml_results)

# Display ML Table
final_ml_results %>%
  gt() %>%
  fmt_percent(columns = Change_Pct, scale_values = FALSE) %>%
  fmt_number(columns = c(Current_Price, Predicted_Price, RMSE), decimals = 2)

Ticker	Current_Price	Predicted_Price	Change_Pct	RMSE
AAPL	248.68	236.12	−5.05%	18.80
ADBE	302.73	306.16	1.13%	8.27
AMZN	239.68	228.56	−4.64%	6.58
BAC	51.58	43.85	−14.99%	6.28
DIS	111.56	110.37	−1.07%	2.68
GOOGL	328.39	188.99	−42.45%	68.56
HD	383.07	379.56	−0.92%	10.66
JNJ	218.29	157.80	−27.71%	27.65
JPM	297.76	259.16	−12.96%	37.55
KO	72.26	67.56	−6.51%	2.29
MA	525.59	524.06	−0.29%	21.06
META	661.78	656.55	−0.79%	39.42
MSFT	469.30	424.53	−9.54%	64.67
NFLX	86.00	86.38	0.44%	19.34
NVDA	187.94	127.50	−32.16%	43.80
PG	150.58	151.22	0.42%	2.29
PYPL	56.91	56.10	−1.43%	1.57
TSLA	448.26	415.91	−7.22%	31.28
UNH	355.18	417.20	17.46%	96.35
V	327.00	328.23	0.38%	9.99
WMT	118.40	97.62	−17.55%	9.10

# Interactive Plotly Chart for the first ticker as an example
plot_ly(data = df_ta %>% filter(ticker == "AAPL")) %>%
  add_lines(x = ~date, y = ~close, name = "Actual Price") %>%
  layout(title = "Stock Price Overview", xaxis = list(title = "Date"), yaxis = list(title = "Price"))